Mike Brock <brockm@...>
Tony Thompson writes:
"Malcolm Laughlin wrote:
I would be very wary of any interptretation of statistics thatIf you haven't read Tim Gilbert's analyses, please do so in the
archives before commenting further."
Well...to be completely fair to all, Malcolm notified me that he wanted to comment on this issue. My response was a suggestion that he read through the many messages regarding this issue during the period 2003-2006 or so. He informed me that he would do that before reaching his conclusions. The point here is that this issue is not a closed subject...anymore than whether life exists on Mars or not. OTOH, it behooves us, I think, to not ignore previously analyzed and presented information...hence, my suggestion to read through the previous messages.
"I found his statistical approach
entirely convincing, and would say, with all due respect, that in my
opinion the burden of proof is on you to show why he was wrong."
Tony's point is well taken. The problem, of course, is that we have so little data. I will add to that that I, myself, have more data that has not been analyzed. My fault. In the case of my analyzed Fraley data, it consists of 34 frt trains in about a month and a half in the spring of '49. Given that UP was running about 35 frt trains [ if I remember correctly ] per day, that's a very tiny bit of data. If the distribution of the approximate number of frt cars moving through Wyoming on a given day, let's say 2800, was uniform, perhaps 80 in a single train is enough to generate a prediction model. Unfortunately, however, the data indicates quite a different sampling. As I pointed out back in June 2006 and repeated on 8-9-2008, the presence or lack thereof of SP box cars was VERY inconsistent. Thus, of the 34 frt trains, 15 trains contained one or less SP box cars and 9 had none at all. This in a population that contained 136 SP box cars. The model of SP box cars being 2.79% of the nation's box cars predicts 46 SP box cars. Not good, an error of 200%. Another significant interchange with UP was C&NW. The model predicted 41 C&NW box cars but the data shows 55...an error of 34%. CB&Q was predicted to be 38. The actual number was 75...an error of 100%. So, what does this mean? Well, consider that the UP Wyoming trunk line interchange with other RR's at Ogden [ SP and D&RGW/WP ], CB&Q [ C&S ] at Cheyenne and Grand Island, NE, and more RR's than I can count at Omaha. Note that the SP line from Sacramento essentially had no interchange between there and Ogden. We're talking about a rather lengthy system with very limited interchange with other RR's. If the UP Wyoming trunk line had numerous interchanges throughout its distance, perhaps like just about any midwestern RR, maybe the population of frt cars WOULD follow the model's predictions. Who knows.
Incidentally, it has been argued that the model predicts over a long period of time...say a yr...rather than just 34 trains in a month and a half. Perhaps, but that doesn't work for me. Every day is May 14, 1953. Do groundhogs live in Wyoming? At least I don't use a clock radio. And, unlike the weather guy, I like to watch big steam power working up the Hill every day...although I will admit to a certain amount of annoyance when the occasional turbine or diesel rolls by.